Video Coding and HEVCmile-high.video/files/mhv2018/pdf/day1/1_01_Grois.pdf · Team on Video Coding (JCT-VC) and issued joint call for proposals (CfP) on video coding technology in

Mile High Video – Denver, CO 2018July 31, 2018

Video Coding and HEVC

Dr. Dan Grois, E-mail: [email protected]

MILE HIGH VIDEO – DENVER 2018

@ Prof. Masayuki Nakajima, Uppsala University


Tutorial Agenda

Part I: Introduction to block based hybrid coding

Part II: Brief Overview of H.264/MPEG-4 AVC

Part III: HEVC version 1 (and version 2)

Part IV: Future Video Coding Development


Part I: Introduction to block-based hybrid video coding

Video Compression Goal:

Efficiently condense visual data, while

Minimizing the loss of visual quality due to the compression.

Why do we need to compress video?

To reduce the storage space;

To reduce the transmission/delivery bit-rate...


Video Coding, Transmission, and Applications

[Grois2015]

Bitstream

~7 MBit/s

~1.2%

Sender Receiver

Channel

HD-Video

uncodiert

~600 MBit/s

HD-Video

uncodiert

~600 MBit/s

HD Video

uncoded

~600 MBit/s

~100%

HD-Video

uncodiert

HD-Video

uncodiert

HD Video

decoded

~600 MBit/s

~100%


Hybrid Video Coding (I)

Block based hybrid

coding:

Division into blocks

Intra/Inter prediction

Transform coding

[Grois2015]


Hybrid Video Coding: Motion Compensation

Current picture Previous picture

[Grois2015]


Hybrid Video Coding: Residual Processing


Residual (difference) picture

after motion compensation

[Grois2015]


Hybrid Video Coding: Residual Processing

Things can be more complicated…


Another previous pictureResidual picture

[Grois2015]


Transform and Quantization of the Residual Picture

Goals:

Adjust bit rate vs. fidelity;

Remove subjectively irrelevant details.

Apply transform (e.g., DCT)

for converting the spatial

domain pixels into transform

domain coefficients;

Apply quantization (e.g.,

scalar) for reducing a number of

levels for transformed

coefficients.

[Grois2015]


[Grois2015]

Standardization organizations

Video Coding Experts Group (VCEG)

Formally structured as Question 6 of ITU-T Study Group 16 (Q.6/SG16)

Informal nickname, VCEG, originated in 1998

History spans from the development of first standards for digital videocoding (ITU-T H.120 and H.261) through recent development of HEVC(ITU-T H.265/MPEG-H Part 2) and its extensions

Also responsible for still-picture coding in partnership with ISO/IEC JTC1/SC 29/WG 1 (JPEG).


[Grois2015]

Standardization organizations

Moving Pictures Experts Group (MPEG)

Formally ISO/IEC JTC 1/SC 29/WG 11 – Coding ofmoving pictures and audio (ISO/IEC Joint TechnicalCommittee 1, Subcommittee 29, Working Group 11)

Established in 1988

Mission to develop standards for coded representationof digital audio and video and related data

Several sub-groups:

Requirements

Systems

Video

Audio

3D Graphics Compression

Test, Communication


H.264/MPEG-4 Advanced Video Coding (AVC)


H.264/MPEG-4 AVC Timeline

Aug. 1999

First test model (TML-1) of H.26L in VCEG

Dec. 2001

Formation of Joint Video Team (JVT) between VCEG and MPEG

JVT Chairs: G. J. Sullivan, A. Luthra, and T. Wiegand

May 2003

ITU-T SG16 Recommendation H.264 approved

International Standard ISO/IEC 14496-10

until April 2004

Extensions Project: Fidelity range extensions (FRExt)

[Grois2015]


H.264/MPEG-4 AVC Common Profiles

Marpe et al., IEEE Comm. Magazine, 2006


H.264/MPEG-4 AVC Summary

Based on hybrid video coding and similar in its virtue to other

standards but with several improvements.

Some new key aspects are:

Enhanced motion compensation;

Small blocks for transform coding;

Improved de-blocking filter;

Enhanced entropy coding.

Significant bit-rate savings of about 50% compared to

H.262/MPEG-2 for the same perceptual quality

(especially, for higher-latency applications allowing B pictures).

[Grois2015]


H.264/AVC Basic Coding Structure

Entropy

Coding

Scaling & Inv.

Transform

Motion-

Compensation

Control

Data

Quant.

Transf. coeffs

Motion

Data

Intra/Inter

Coder

Control

Decoder

Motion

Estimation

Transform/

Scal./Quant.-

Input

Video

Signal

Split into

Macroblocks

16x16 luma pels

Intra-frame

Prediction

Deblocking

Filter

Output

Video

Signal

[Grois2015]


Common Elements with Previous Standards

Macroblocks: 16x16 luma + 2 x 8x8 chroma samples;

Input: Association of luma and chroma and

conventional sub-sampling of chroma (4:2:0);

Block motion displacement;

Motion vectors over picture boundaries;

Variable block-size motion;

Block transforms;

Scalar quantization;

I, P, and B coding types.

[Grois2015]


Macroblocks and partitions

Entropy

Coding

Scaling & Inv.

Transform

Motion-

Compensation

Control

Data

Quant.

Transf. coeffs

Motion

Data

Intra/Inter

Coder

Control

Decoder

Motion

Estimation

Transform/

Scal./Quant.-

Input

Video

Signal

Split into

Macroblocks

16x16 luma pels

Intra-frame

Prediction

De-blocking

Filter

Output

Video

Signal

Motion vector accuracy 1/4 (6-tap filter)

8x8

0

4x8

0 10 1

2 3

4x48x4

1

08x8

Types

0

16x16

0 1

8x16

MBTypes

8x8

0 1

2 3

16x8

1

0

[Grois2015]


Multiple Reference Frames and Generalized Bi-Predictive Frames

Current

Picture4 Prior Decoded Pictures

as Reference

1

Extend motion vector

by reference picture

index .

Provide reference

pictures at decoder

side.

In case of bi-predictive

pictures: decode 2

sets of motion

parameters.

Can jointly exploit scene cuts, aliasing, uncovered

background and other effects with one approach

3

0

[Grois2015]


Weighted Prediction

In addition to shifting in spatial position, and selecting fromamong multiple reference pictures, each region’sprediction sample values can be:

Multiplied by a weight; and

Given an additive offset.

Some key uses:

Improved efficiency for B coding, e.g.: acceleratingmotion, multiple non-reference B temporally betweenreference pics.

[Grois2015]


Intra Prediction

Entropy

Coding

Scaling & Inv.

Transform

Motion-

Compensation

Control

Data

Quant.

Transf. coeffs

Motion

Data

Intra/Inter

Coder

Control

Decoder

Motion

Estimation

Transform/

Scal./Quant.-

Input

Video

Signal

Split into

Macroblocks

16x16 luma pels

Intra-frame

Prediction

De-blocking

Filter

Output

Video

Signal

Directional spatial prediction

(9 types for luma, 1 chroma)

1

2

3456

7

8

0

Q A B C D E F G H

I a b c d

J e f g h

K i j k l

L m n o p

[Grois2015]


Transform Coding

Entropy

Coding

Scaling & Inv.

Transform

Motion-

Compensation

Control

Data

Quant.

Transf. coeffs

Motion

Data

Intra/Inter

Coder

Control

Decoder

Motion

Estimation

Transform/

Scal./Quant.-

Input

Video

Signal

Split into

Macroblocks

16x16 pixels

Intra-frame

Prediction

De-blocking

Filter

Output

Video

Signal

4x4 Block Integer Transform

Repeated transform of DC coeffs

for 8x8 chroma and some 16x16

Intra luma blocks

1 1 1 1

2 1 1 2

1 1 1 1

1 2 2 1

H

[Grois2015]


Entropy Coding

Entropy

Coding

Inv. Scal. &

Transform

Motion-

Compensation

Control

Data

Quant.

Transf. coeffs

Motion

Data

Intra/Inter

Coder

Control

Decoder

Motion

Estimation

Transform/

Scal./Quant.-

Input

Video

Signal

Split into

Macroblocks

16x16 luma pels

Intra-frame

Prediction

De-blocking

Filter

Output

Video

Signal

[Grois2015]


Binary Coding: CAVLC

Context-Adaptive Variable Length Coding (CAVLC)

Exp-Golomb code for all symbols except for transform coefficients

Context adaptive VLCs for coding of transform coefficients

No end-of-block, but number of coefficients is decoded;

Coefficients are scanned backwards;

Contexts are built dependent on transform coefficients.

[Grois2015]


Binary Coding: CABAC

Context-based Adaptive Binary Arithmetic Coding (CABAC)

Usage of adaptive probability models for most symbols;

Exploiting symbol correlations by using contexts;

Restriction to binary arithmetic coding:

Simple and fast adaptation mechanism;

Fast binary arithmetic codec based on table look-ups and shifts only.

[Grois2015]


Part II: High Efficiency Video Coding Version 1



A Guide to the H.265/HEVC Standard and its Extensions

Cambridge University Press.

Printed Fully In Color.

Includes HEVC Extensions.

Coming in 2019.


Motivation for Improved Video Compression

History of Video Coding Standards

Jevons Paradox

"The efficiency with which a resource is used

tends to increase (rather than decrease) the

rate of consumption of that resource."

[Wikipedia, Bross2016]

William Stanley Jevons


Motivation for Improved Video Compression

IP video traffic will be 82% of all consumer Internettraffic by 2021 [Cisco2017].

For enabling High-Quality video services, efficientcompression techniques are required, especially for3840x2160 (4K) or 7680x4320 (8K) resolutions.

http://moneytipcentral.comhttp://www.theexpgroup.com


H.265/HEVC – Applications

Internet streaming, download and play

Real-time conversational services

Broadcast

Mobile streaming, conversational services and broadcast

Content production and distribution

Home and Digital Cinema

Camcorders

Medical imaging

Remote video surveillance

Storage media (e.g., disks, digital video tape recorder)

Wireless display


HEVC and the JCT-VC Partnership

ITU-T VCEG and ISO/IEC MPEG established Joint Collaborative Team on Video Coding (JCT-VC) and issued joint call for proposals (CfP) on video coding technology in 2010.

As a result, there was an intensive development of the so-called High-Efficiency Video Coding (HEVC) standard during the next two and the half years.

2013: HEVC version 1;

2014: HEVC version 2 – Range Extensions (RExt), Scalable Extensions (SHVC), Multiview Extensions (MV-HEVC);

2015: HEVC version 3 – 3D Video Coding Extensions (3D-HEVC);

2016: HEVC version 4 – Screen Content Coding Extensions (HEVC-SCC);

2018: HEVC version 5 – additional SEI messages that include omnidirectional video SEI messages, a Monochrome 10 profile, a Main 10 Still Picture profile.

http://mpeg.chiariglione.org/


Scope of Video Coding Standardization

Only restrictions on bitstream (syntax & semantics), and decoding process are standardized:

Permits optimization beyond the obvious;

Permits complexity reduction for implementations;

Provides no guarantees of quality.

Pre-Processing EncodingSource

Destination

Post-Processing

& Error RecoveryDecoding

Scope of a Standard

[Grois2015]


History: Call for Proposals Testing

27 complete proposals submitted (some multi-organizational);

Each proposal was a major package – lots of encoded video, extensive documentation, extensive performance metric submissions, sometimes software, etc;

Extensive subjective testing (3 test labs, 4 200 video clips evaluated, 850 human subjects, 300 000 scores);

Quality of the proposal video was compared to H.264/MPEG-AVC anchor encodings;

In a number of cases, comparable quality at half bit rate;

Test report issued as document JCTVC-A204.

[Grois2015]


History: Call for Proposals Results

All proposals basically conceptually similar to H.264 / AVC (and prior standards):

Block-based with variable block sizes;

Block motion compensation;

Fractional-pel motion vectors;

Spatial intra prediction;

Spatial transform of residual difference;

Integer-based transform designs;

Arithmetic or VLC-based entropy coding;

Various methods of in-loop filtering to form final decoded picture and improve prediction.

"Test Model under Consideration" was set up from common design elements of well-performing proposals (Document JCTVC-A205).

[Grois2015]


History: JCT-VC Meetings

April 2010 to January 2013 – 12 JCT-VC Meetings

HEVC version 1 was officially finalized in January 2013.

0

200

400

600

800

1000

1200

Participants

Documents

First Test Model under

Consideration (TMuC)

First Working Draft and

Test Model (Draft 1 / HM1)

ISO/IEC CD

(Draft 6 / HM6)

ISO/IEC DIS

(Draft 8 / HM8)

ISO/IEC FDIS &

ITU-T Consent

(Draft 10 / HM10)

[Grois2015]


Design Overview

• High-level design similar to H.264/MPEG-4 AVC:

Scaling & Inverse Transform

Intra-Picture Prediction

Inter-Picture Prediction

Entropy Coding

Output

010110...

Bitstream

Transform, Scaling & Quantization

Motion Estimation

Decoder

-

Video coding layer (VCL):

Coded representation of the

picture samples

Hybrid video coding

Network abstraction layer (NAL):

Packet oriented payload

NAL payload types for parameter

sets (header information) and data

of coding layer

Pictures structured into slices, with

additional features supporting

parallel processing (tiles,

wavefront) and improved

packetization granularity

(dependent slices)

[Grois2015]


Coding-Layer Design

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream

Quant. Transf. Coeffs.

Transform,

Scaling & Quantization

Output

Video Signal

Motion Estimation

Decoder

-Subdivision

into Blocks

Input

Video Signal

[Grois2015]


Block Structures

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream


Transform,


Output

Video Signal

Motion Estimation

Decoder

-Coding

Tree

Blocks

Coding Blocks

Prediction

Blocks

Transform Blocks

Input

Video Signal

[Grois2015]


Block Structures (Cont.)

Variable Block sizes starting from grid of Coding Tree Units (CTU):

• Consisting of Coding Tree Blocks (CTB) with fixed sizes for lumaand chroma;

• Luma CTB size typically 64x64 but can be set to 32x32 and 16x16 (size signaled in SPS).

• Processed in raster scan order.

4 5 6 7 8 9 10

11 12 13 14 15 16 17 18 19 20 21

22 23 24 25 26 27 28 29 30 31 32

33 34 35

60 61 62 63 64 65

66 67 68 69 70 71 72 73 74 75 76

77 78 79 80 81 82 83 84 85 86 87

88 89 90 91 92 93 94 95 96 97 98

49 50 51 52 53 54

55 56 57 58 59

36 37 38 39 40 41 42 43

44 45 46 47 48

0 1 2 3

[Grois2015]



Coding Tree Block is the root of the Coding Quadtree:

A Coding Block leave can be predicted using Inter- or Intra prediction;

The coding quadtree is traversed in Z–Scan Order

14

15 16

17 18

2 3

6 12 13

4 5

0 1 7

8 9

10 11

[Grois2015]



Coding Quadtree:

4K sequence Traffic (cropped to 2560x1600)

Detail showing coding

quadtree with Coding Blocks

Simple scheme to locally

adapt block sizes

Large block sizes especially

for HD resolutions and

beyond

[Grois2015]



A Coding Block can be split:

once into Prediction Block partitions;

recursively into Transform Blocks using the Residual Quadtree (RQT): transform sizes ranging from 4x4 to 32x32.

E.g., Nx2N partition with 2

prediction blocks.

E.g., RQT with depth=3 and

10 transform blocks.

14

15 16

17 18

2 3

6 12 13

4 5

0 1 7

8 9

10 11

0 1

0 1

2 34

7

5 6

8 9

[Grois2015]



Example:

[Schwarz2014]

[Grois2015]



Different Prediction Block partition modes:

Each prediction block in a coding block uses the same prediction mode (intra or inter);

Inter: motion compensation is done per prediction block and prediction blocks can be “merged”;

Intra: mode, e.g. DC, planar, angular,.. defined per prediction block.

2NxnU 2NxnD nLx2N nRx2N2NxN Nx2N

Inter prediction only

2Nx2N NxN

Asymmetric motion partitioning (AMP)

[Grois2015]



RQT with variable-size Transform Block leaves:

Transform of the residual is performed on transform blocks;

Intra prediction is also performed on transform blocks in intra coding blocks;

Transform sizes ranging from 4x4 to 32x32;

Adaptation to varying space-frequency characteristics in the residual signal for the DCT-based integer transform.

[Grois2015]


Average bit-rate savings for successively increasing the CTU size

and the number of hierarchy levels in the coding tree:

Anchor: 16x16 CTU size and a minimum CU size of 8x8 luma samples.


CTU Size and Minimum CU SizeEntertainment

Applications

Interactive

Applications

32x32 - 16x16 9.2% 17.4%

32x32 - 8x8 12.1% 20.2%

64x64 - 16x16 12.7% 23.8%

64x64 - 8x8 14.9% 25.5%

[Schwarz2014]

[Grois2015]


Coding efficiency improvement for successively increasing the

maximum TU size:

Anchor: 4x4 TUs, 64x64 CTU, all CU and PU sizes


Maximum TU SizeEntertainment

Applications

Interactive

Applications

Maximum TU Size of 8x8 6.8% 8.5%



[Schwarz2014]

[Grois2015]



Average bit-rate savings for successively enabling various

PU sizes:

Anchor: 16x16 PUs and 4x4 TUs[Schwarz2014]

Enabled PU SizesEntertainment

Applications

Interactive

Applications

16x16 and 8x8 2.7% 4.4%

Square PUs from 4x4 to 16x16 6.1% 5.6%



All modes except asym. (+4x4 PUs) 20.0% 31.0%

All HEVC PU Sizes (+4x4 PUs) 20.7% 33.0%

[Grois2015]


Intra-Picture Prediction

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream


Transform,


Output

Video Signal

Motion Estimation

Decoder

-

Input

Video Signal

[Grois2015]


Intra-Picture Prediction: Main Tools

H.264/AVC intra-picture prediction drawbacks:

Insufficient range of supported coding block sizes: especially

high-resolution videos (HD or UHD) cannot be coded properly due to

the poor representation of some picture textures.

Insufficient number of intra prediction directions: especially for

larger block sizes.

Insufficient prediction of homogeneous regions: visible artifacts

introduced by discontinuities at block boundaries

Inconsistency across block sizes: depending on the size of the

block, H.264/AVC uses different methods for predicting a block and

the color component, which is represented by the block; it turn, this

can be especially critical for a large variety of block sizes.

[Grois2015]


Intra-Picture Prediction: Main Tools

HEVC intra-picture prediction addresses H.264/AVCdrawbacks by:

Reference samples computed with 1/32 pel accuracy (bilinear interpolation);

Angular prediction with 33 directional modes as well as Planar and DC modes;

Reference sample smoothing for diagonal directions and planar mode;

Boundary smoothing (across block boundary) for the horizontal, vertical and DC modes;

Prediction modes unified across all block sizes.

[Grois2015]


Three prediction types:

Planar prediction (0),

DC prediction (1),

33 angular prediction directions (2-34).

Intra Prediction: Main Tools (Cont.)

[Grois2015]


Intra Prediction: Main Tools (Cont.)

Only 5 modes for chroma can be signaled:

Planar (0);

DC (1);

Horizontal (10);

Vertical (26);

Same directional mode as luma.

[Grois2015]


Intra Prediction: Reference Sample Smoothing

HEVC adaptive smoothing depends on:

Block size;

Directionality;

Amount of detected discontinuity.

Two smoothing filters:

Normal filtering;

Strong filtering.

Applied to each reference sample using the neighboring reference samples.

[Grois2015]


Intra Prediction: Reference Sample Smoothing (Cont.)

Reference Sample Smoothing depends on:

Block size;

Intra direction;

Discontinuity by comparing minimal distance of direction to horizontal and vertical mode with predefined thresholds

[Grois2015]


Intra Prediction: Boundary Sample Smoothing

Discontinuities along block boundaries may be introduced by applying intra-prediction modes.

Hence, HEVC employs boundary sample smoothing for:

Luma TBs;

TB sizes < 32x32;

Horizontal prediction mode, applied to the first row;

Vertical prediction mode, applied to the first column;

DC intra-prediction mode, both the first row and column of samples in the TB are filtered.

Prediction for chroma components tends to be very smooth -> no boundary smoothing for chroma TBs to reduce the computational complexity

[Grois2015]


Inter-Picture Prediction

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream


Transform,


Output

Video Signal

Motion Estimation

Decoder

-

Input

Video Signal

[Grois2015]


HEVC improves motion compensated prediction by:

Asymmetric Motion Partitions (AMP) as explained in Block Structures;

Reference Picture Sets (RPS) for simplified decoded picture buffer

management;

Improved interpolation filters for luma and chroma;

More efficient motion data coding including

Advanced Motion Vector Prediction (AMVP);

Inter-picture prediction block merging (Merge);

Motion Data Storage Reduction (MDSR);

Simplified weighted sample prediction.

No 4x4 blocks and only uni-prediction for 4x8/8x4 to reduce memory

bandwidth

Inter-Picture Prediction: Main Tools

[Grois2015]


Transform and Quantization

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream


Transform,


Output

Video Signal

Motion Estimation

Decoder

-

Input

Video Signal

[Grois2015]


Transform and Quantization

Prediction residual is transformed (unless signaled as zero by coded block flag, skip mode or alternative bypass modes);

Core transforms 4x4 ... 32x32 are integer approximations of DCT:

Close to orthogonality;

Nearly identical norms of all basis vectors: Unique scaling factor for each transform size;

Matrices of smaller-size transforms are sub-sampled versions of length-32 transform matrix.

2D transforms separable (except for the rounding after each transform step), only square transforms used.

[Grois2015]


Transform and Quantization (Cont.)

925435770808790908780705743259

18507589897550181850758989755018

257090804395787875794380907025

36838336368383363683833636838336

439057258770980809708725579043

50891875751889505089187575188950

578025909874370704387990258057

64646464646464646464646464646464

704387990258057578025909874370

75188950508918757518895050891875

809708725579043439057258770980

83363683833636838336368383363683

875794380907025257090804395787

89755018185075898975501818507589

908780705743259925435770808790

64646464646464646464646464646464

H

• Example: length-16 transform matrix (Euclid. norm approx. 256):

8x8

[Grois2015]



• Example: length-16 transform matrix (Euclid. norm approx. 256):

925435770808790908780705743259

18507589897550181850758989755018

257090804395787875794380907025

36838336368383363683833636838336

439057258770980809708725579043

50891875751889505089187575188950

578025909874370704387990258057

64646464646464646464646464646464

704387990258057578025909874370

75188950508918757518895050891875

809708725579043439057258770980

83363683833636838336368383363683

875794380907025257090804395787

89755018185075898975501818507589

908780705743259925435770808790

64646464646464646464646464646464

H

4x4

[Grois2015]



Integer approximation of Discrete Sine Transform (DST) is applied to 4x4 intra residual:

In intra prediction (in particular directional and planar), the prediction error increases with larger distance from the boundary;

DST basis vectors are better fitting such behavior;

For intra residual of 4x4 PBs, the following transform matrix is used (Euclid. norm approx. 128, same as length-4 DCT (the Euclidean norm is the square root of the sum of all the squares)).

Provides a 1% bit-rate reduction but was restricted to Intra 4×4 luma transform blocks due to marginal benefits for the other transform cases.

29 55 74 84

74 74 0 74

84 29 74 55

55 84 74 29

H

[Grois2015]



Quantization similar to H.264/MPEG-AVC (QP+6 = double step size, QP values are defined from 0 to 51 ).

The QP to Qstep mapping has a reduction of about 12.5% in the bitrate for the increase of 1 in QP.

CU level adaption of quantizer step size.

Coding of QP performed differentially within CTU.

Frequency weighted quantization possible via quantization matrices (encoded in SPS or PPS):

For the larger transforms (16x16, 32x32) a subsampled quantization matrix of size 8x8 is used, i.e. same quantization step sizes used for groups of 2x2 or 4x4 adjacent coefficients);

Different quantization matrices for intra and inter, Y/Cb/Cr;

Matrix entries coded differentially.

[Grois2015]


In-loop Filters

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream


Transform,


Output

Video Signal

Motion Estimation

Decoder

-

Input

Video Signal

[Grois2015]


Two filters to remove coding artifacts and preserve edges applied before writing the output to reference memory (not effective in intra prediction) :

Deblocking Filter (similar to H.264/MPEG-4 AVC, but having a reduced number of filtering strengths and better parallelization) –operated only on 8x8 block boundaries (not 4x4) with 4-sample units.

Sample Adaptive Offset (SAO) - two different offset methods:

Individual mapping with edge offset values applied to

samples based on classification as an edge or non-edge.

Non-linear amplitude mapping function with band offset

values applied to samples based on their values lying in one of

four consecutive samples value ranges (bands).

In-loop Filters

[Grois2015]


Deblocking Filter

Generally, when designing a deblocking filter, several issues should be carefully considered:

Filter has a direct impact on the subjective picture quality.

The filtering decisions contain challenges that lead to the following main questions:

Is the block boundary natural edge or artifact (filter off/on)?

Which filtering strength should be used?

[Grois2015]


Deblocking Filter (Cont.)

The deblocking is performed on a four-sample portion of a 8x8 block boundary, when the following conditions are held:

the block boundary is a prediction unit or transform unit boundary;

the block boundary strength, which is determined according to a predefined criteria, is greater than zero;

variation of signal on both sides of a block boundary is below a specific threshold.

[Grois2015]



Splitting a picture on 8x8. The dotted-lines represents non-overlapping blocks of 8x8 samples: these non-overlappingblocks can be deblocked in a parallel manner betterparallelization.

[Grois2015]



• BS values for the boundary between two neighbor lumablocks

• 2 – Strong Filtering, 1 – Weak Filtering, 0 – No Filtering:

[Grois2015]


Deblocking: Weak/Strong Filtering

The strong filtering is applied:

on three samples on both sides of the block boundary;

within the four-sample segment;

similar to H.264/MPEG-4 AVC.

The weak filtering is applied:

on two samples on both sides of the block boundary;

within the four-sample segment.

[Grois2015]


Deblocking Filter: Parallelism

The deblocking process in HEVC provides improved parallelization, compared to H.264/MPEG-4 AVC:

the deblocking process of non-overlapping 8x8 samples can be fully parallelized;

any vertical block edge can be deblocked in parallel with the deblocking of any other vertical block edge;

any horizontal block edge can be deblocked in parallel with the deblocking of any other horizontal block edge;

the updated sample values after the deblocking of vertical block boundaries are used as input for filtering the horizontal block boundaries.

[Grois2015]


Sample Adaptive Offset: Motivation

Sample Adaptive Offset filter is designed to add correctiveoffset values for attenuating:

Systematic Errors introduced by quantization and phase shiftsfrom inaccurate motion vectors;

Ringing Artefacts (Gibbs Phenomenon), introduced mainlyby large transform sizes.

Two different offset methods can be used:

Non-linear amplitude mapping function with band offsetvalues applied to samples based on their values lying in one offour consecutive samples value ranges (bands) out of 32;

Individual mapping with edge offset values applied tosamples based on classification as edge or non-edge, whereasthe gradient orientation is determined by the encoder.

[Grois2015]


Entropy Coding

Scaling &

Inverse Transform

Intra-Picture

Prediction

Inter-Picture

Prediction

Entropy

Coding

In-Loop Filter

Output

010110...

Bitstream


Transform,


Output

Video Signal

Motion Estimation

Decoder

-

Input

Video Signal

[Grois2015]


Context-based Adaptive Binary Arithmetic Coding

Single entropy coding scheme in HEVC using Context-based Adaptive Binary Arithmetic Coding (CABAC) instead of two in H.264/AVC:

Usage of adaptive probability models for most symbols;

Exploiting symbol correlations by using contexts;

Restriction to binary arithmetic coding based on table look-ups and shifts only;

Main improvements, compared to H.264/AVC:

Reduced number of contexts;

Reduced number of context coded binary decisions to avoid throughput bottleneck;

Removed dependencies to facilitate parallel context derivation;

Truncated Rice Codes added as binarization.

[Grois2015]


Entropy Coding of Transform Coefficients

Sub-blocks and the 16 coefficients in a sub-block are processed usinga reverse diagonal scan.

Intra 4x4 TBs and 8x8 luma TBs use a horizontal scan and avertical scan.

8 s

am

ple

s

horizontal

horizontal

vert

ica

l

16 samples

16

sa

mp

les

4 samples

4 sa

mp

les

diagonal diagonal

8 sa

mp

les

vertical3

2

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18 19 20 21 22 23 24 25 2726 28 29 30 31 32 33 34

[Grois2015]


Alternative Coding Modes (I)

Three special coding tools give support for special cases

(without any significant implementation cost).

Transform bypass

Lossless bypass

Intra PCM Mode

[Grois2015]


Alternative Coding Modes (II)

Transform bypass:

Signaled at TB level (transform_skip_flag);

Useful particularly for graphics content (sharp edges, flat areas),

e.g. as in screen content coding applications;

CABAC contexts of transform coding are directly applied to

prediction error signal (becoming spatial contexts to close

neighbor samples).

Bypass enabling "lossless" coding:

Signaled at CU level (cu_transquant_bypass_flag);

Bypass of scaling, transform and in-loop filters;

CABAC entropy coding still applied as is;

Can also be used locally, e.g. for graphics/text[Grois2015]


Alternative Coding Modes (III)

Intra PCM mode:

Signaled at CU level (pcm_flag);

Direct coding of sample values;

Is intended for "emergency" situations when encoder:

faces complicated content and would produce more bit rate than

uncoded samples (e.g., in case of the uncorrelated noise);

would overrun its processing capabilities.

[Grois2015]


Alternative Coding Modes (IV)

[Grois2015]


High-Level Syntax

• In general, all syntax elements above the slice segment data layer are called high-level syntax.

• High-level syntax:

Access to packets

Settings of low level coding tools

Random-access information

Metadata

[Grois2015]


High-Level Syntax: Parameter Sets

Sequence Parameter Set (SPS) and Picture Parameter Set (PPS) similar as H.264/MPEG-AVC

New Video Parameter Set (VPS)

Describes overall characteristics of coded video sequences, including the dependencies between layers.

Enables compatible extensibility in terms of signaling at the systems layer.

[Grois2015]


High-Level Syntax: VPS

The (HEVC version 1) Video Parameter Set contains:

Maximum number of layers in the bitstream and the highest layer ID

Maximum number of temporal sub-layers

Profile/Tier/Level (for each temporal sub-layer)

DPB parameters (optional for temporal sub-layers)

Layer sets (which can be extracted as sub-bitstreams)

Timing (optional)

HRD parameters (optional)

[Grois2015]


High-Level Syntax: VPS (Cont.)

The Video Parameter Set extension (HEVC version 2)contains:

Scalability types

Scalable

Multi-View

Depth (v3)

Auxiliary pictures

Dependencies between layers

Output layer sets

Representation Format

DPB information

VPS VUI

[Grois2015]


High-Level Syntax: SEI and VUI

Supplemental Enhancement Information (SEI) and Video Usability Information (VUI):

SEI and VUI concepts similar to H.264/MPEG-AVC

no impact on normative decoder behavior

usage optional

Example SEI messages: Buffering Information,Picture Timing, Frame packing (for stereo), display orientation, interlaced-scan indicators (frame/field arrangement);

Example VUI parameters: Hypothetical Reference Decoder (HRD) and clock parameters, color space,aspect ratio, constraint flags.

[Grois2015]


Parallel Picture Processing

Slices / Slice segments

Tiles

Wavefront Parallel Processing (WPP)

Parallelization Tools

[Grois2015]


Slice coding:

Slices can be coded independent of each other.

In-loop filters can be disabled at slice boundaries.

Dependent slice segments

Inherit the previous entropycoding state

Parallelization Tools: Slices

[Grois2015]


Tiles are:

Independent rectangular regions of the picture

In a regular pattern(division by horizontal andvertical lines within thepicture)

Processing of tiles

In raster scan order over the tiles

and CTU-wise in raster scan order within the tiles.

Parallelization Tools: Tiles

[Grois2015]


Wavefront processing:

Allows to run several processing threads in a slice over rows of

CTUs with a delay that allows adaptation;

CABAC contexts are inherited from a row above

Parallelization Tools: Wavefronts

[Grois2015]


Parallelization Tools: Performance

[Chi2012]

Total coding losses (using weighted BD-Rate) of different tile configurations for 1080 videos.

[Grois2015]


Parallelization Tools: Performance

Comparison of Parallelization Approaches

[Chi2012]

Properties Slices Tiles WPP

Coding loss Very high High Low

Boundary artifacts Yes Yes No

Single-core issues No Yes No

Parallel scalability Medium Medium Medium/high

Region of interest No Yes No


Onion-like structure

Profiles, Levels and Tiers

Main Still PictureMain 10

+ 8-10 bit

sample bit depth

Main

• Intra coding tools,

• Wavefront parallelism only when

multiple tiles in a picture are not used

• Minimum luma tile size 256x64

• 4:2:0 chroma sampling

• 8 bit sample bit depth

+ Inter

coding tools

+ Timing

information /

constraints

[Grois2015]


Rate-Distortion Performance

Interactive Applications (Low Delay)

Entertainment Applications (Random Access)

[Ohm2012]


Average Bit-rate Savings (Still Picture Coding)

Average bit-rate savings for the HEVC Main Still Picture Profile:

[Nguyen2015]


Average Bit-rate Savings (Still Picture Coding)

Average bit-rate savings relative to JPEG:

[Nguyen2015]


Subjective Verification Tests of HEVC vs. AVC

Verification test completed in April, 2014 (JCTVC-Q1011):

A subjective evaluation was conducted comparing the HEVC Main profile to the AVC High profile;

20 test sequences: 480p to Ultra HD (UHD) various bit rates/quality levels;

Average bit-rate savings for test sequences:

•UHD - 64%

•1080p - 62%

•720p - 56%

•480p - 52%

[Grois2015]


Average Bit-Rate Savings (Random Access)

PCS 2013:


Average Bit-Rate Savings (Random Access)

[Grois2013]

PCS 2013: Rate-Distortion plot examples


Average Bit-Rate Savings (Random Access & Low Delay)

HEVC vs. H.264/AVC High Profile (x264) and VP9

PCS 2013:

SPIE 2014:


Fixed QP Test Case

[Grois2017]

BD-BR: Weighted PSNRYUV

AV1 and VP9 with Rate Control Disabled(negative BD-BR values indicate actual bit-rate savings)

anchor te

st c

an

did

ate

AV1 JEM VP9 HM

AV1 111.8% -17.1% 47.7%

JEM -51.4% -62.0% -29.8%

VP9 21.0% 173.7% 92.5%

HM -30.6% 43.4% -46.6%

SPIE 2017:


Rate Control Test Case

[Grois2017]

anchor

test

ca

nd

ida

te

AV1 JEM VP9 HM

AV1 55.0% -20.0% 9.5%

JEM -34.8% -47.3%

VP9 28.5% 92.2% 37.9%

HM -7.8% -27.0%


AV1 and VP9 with Rate Control Enabled (negative BD-BR values indicate actual bit-rate savings)

SPIE 2017:


Fixed QP Test Case

CODECS JEM HM AV1 VP9

JEM -31.6% -45.6% -58.8%

HM 47.4% -21.7% -40.3%

AV1 89.3% 30.5% -23.4%

VP9 154.8% 73.5% 31.1%



[Nguyen2018]

PCS 2018:


Fixed QP Test Case

[Nguyen2018]


JEM 8.48 X 0.26 X 15.2 X

HM 0.12 X 0.03 X 1.79 X

AV1 3.83 X 32.45 X 58.16 X

VP9 0.07 X 0.56 X 0.02 X

Encoders Run Time


PCS 2018:




JEM -31.6% -32.0% -48.7%

HM 47.4% -1.4% -25.0%

AV1 48.6% 2.3% -22.9%

VP9 97.9% 34.5% 30.2%


AV1 and VP9 with Rate Control Enabled(negative BD-BR values indicate actual bit-rate savings)

[Nguyen2018]

PCS 2018:




JEM 8.48 X 0.40 X 22.58 X

HM 0.12 X 0.05 X 2.66 X

AV1 2.47 X 20.95 X 55.82 X

VP9 0.04 X 0.38 X 0.02 X

Encoders Run Times

AV1 and VP9 with Rate Control Enabled(negative BD-BR values indicate actual bit-rate savings)

[Nguyen2018]

PCS 2018:


HEVC version 2

Format Range Extensions (RExt)

Multilayer Extensions:

Scalability Extension (SHVC)

Multiview Extension (MV-HEVC)


RExt: Motivation

Extent 4:2:0 8-10 bit consumer oriented scope of HEVC version 1 by:

high quality distribution in broadcast, 4:2:0, 12 bit

contribution in broadcast, 4:2:2, 10/12 bit

production and high fidelity content acquisition, 4:4:4, 16 bit, R’G’B’, high

bit rate

medical imaging, 4:0:0 monochrome, 12-16 bit, (near) lossless

alpha channels and depth maps, 4:0:0 monochrome, 8-bit

high quality still pictures, 4:4:4, 8-16 bit, arbitrary picture size

and many others...

[Grois2015]


RExt: Modifications of HEVC version 1

Three categories:

1. Necessary modifications to extend support for chroma formats beyond 4:2:0

and bit depth beyond 10 bits per sample.

2. Coding efficiency improvements for extended formats, lossless and near

lossless coding by means of:

Modified HEVC v1 tools;

New tools:

Adaptive chroma QP offset (ACQP)

Cross Component Prediction (CCP)

Residual Delta Pulse Code Modulation (RDPCM)

3. Precision and throughput optimizations for very high bit rates and bit depths.[Grois2015]


RExt (HEVC Version 2): Profiles

21 profiles added resulting from combination samples of

Prediction type (Intra/Inter);

Chroma format;

Bit-depth;

Tool set.

4:0:0

4:2:0

4:2:2

4:4:4

s

8 10 12 16

Chr

oma

Form

at

Bit-Depth

Intra Profiles

S

S

S

8 10 12 16

Bit-Depth

Inter ProfilesIncluding High Throughput profile

SIncluding equivalent Still Picture profile (SPP)

Version 1 profile including SPP

1 1

1

Adaptive Chroma QP

RDPCM, CCP, Rice and TSM/TQB modifications

Extended Precision

Bypass Alignment is only in the High

Throughput profile

[Flynn2015]


RExt: Profiles (Cont.)

[Grois2015]


SHVC: Motivation

[Grois2015]

Traditional approach:

Encode and store a separate video stream for all

possible clients and connection speeds

Scalable Video Coding:

Encode a common base layer

Add enhancement layers as needed


SHVC: Motivation (Cont.)

[Grois2015]

Video streaming with multiple layers

Images by Freepik.com

LTE

3G

Wifi

DSL

Base layer

Quality

enhancement

Resolution

enhancement #1

Resolution

enhancement #2

http://www.freepik.com/free-photos-vectors/background


SHVC: Advantages and disadvantages of scalable coding

[Grois2015]

Generic advantages:

Less storage space

Less data rate in multi- and broadcast environments

(Less encoding time)

More flexibility (i.e. combinations of layers)

Built-in support for up- and down-switching

Error resilience: Unequal error protection

Generic disadvantages:

Overhead compared to single layer

Higher decoder complexity


SHVC: Single- vs. multi-loop functionality

[Grois2015]

Single-loop decoder (e.g. SVC):

1. Parse base layer

2. Parse enhancement layer, using information from base layer

3. (decode base-layer intra blocks)

4. Decode enhancement layer

One decoding process overall

Multi-loop decoder (e.g. SHVC):

1. Decode base layer

2. Upsample result as reference signal

3. Decoder enhancement layer

One decoding process per layer


SHVC: Properties of single- vs. multi-loop

[Grois2015]

Single-loop decoder (e.g. SVC):

Low decoding complexity

Always requires changes in low-level coding tools

Minimal re-use of existing components

Multi-loop decoder (e.g. SHVC):

Higher decoding complexity

May not need changes in low-level coding tools

Allows re-use of existing single-layer designs as base for

each layer-decoder


SHVC: Scalability types (I)

[Grois2015]

Temporal scalability (HEVC version 1)

Higher frame rates in enhancement layer

e.g. 30 fps base layer to 60 fps enhancement layer

Spatial scalability

Higher spatial resolutions in enhancement layer

e.g. 720p base layer to 1080p enhancement layer

Coarse grain SNR scalability

Higher SNR qualities in enhancement layers

e.g. low-quality base-layer at 1 MBit/s to high-quality enhancement-

layer at 8 MBit/s

External base layer scalability (new)

Base layer encoded by another external encoder

e.g. H.264/AVC base layer with SHVC enhancement layer


SHVC: Scalability types (II)

[Grois2015]

Bit depth scalability (new)

Higher bit depths in enhancement layer

e.g. base layer with bit depth of bit 8 to enhancement layer with bit

depth of 10 bit

Interlace-to-progressive scalability

Base layer in interlace format, enhancement layer in progressive

format

Color gamut scalability (new)

Higher color gamut in enhancement layers

e.g. BT.709 color gamut in base layer to BT.2020 in enhancement

layer


MV-HEVC: Motivation

[Grois2015]

Use cases

Stereo (2-view) “3D” movies and TV programs

Broadcast

Distributed media (Blu-Ray)

On-demand streaming

Cinema

Multiple view displays

Alternatives

Frame-compatible formats

Simulcast

MV has to provide better performance.


MV-HEVC: Design

[Grois2015]

High-Level Syntax only

Similar to MVC

Takes advantage on multi-layer syntax elements in

HEVC design

No changes at block-level

Allows re-use of existing HEVC encoder and decoder

components

Only changes in decoded picture buffer (DPB) and

reference picture list creation


“The outlook for HEVC use in the digital video market is

extremely positive...” [DTC2016]

HEVC Products Forecast Overview


Future Video Coding Development

What is Next?

Mile High Video – Denver, CO 2018July 31, 2018 122

The work on the video coding techniques beyond HEVC started already in 2015:

• Joint Video Exploration Team (JVET) of MPEG and VCEG organizations

was established last October, 2015, in Geneva [Bross2016].

• JEM: Joint Video Exploration Model Software.

Future video coding standard should consider from the beginning

(among others) [Bross2016]:

• UHD: 4K and up

• High Dynamic Range (HDR) and Wide Color Gamut content

• High Frame Rates

• High-Level Scalability (like SHVC)

• Drone Video

• 360 Video



List of several coding tools that improve coding efficiency of HEVC: [SG16−C0806]

Larger coding tree blocks and larger transforms:

The CTU size signaled in the sequence level is set to be 256x256 by

default;

Supporting Coding Tree Units (CTUs) larger than 64x64;

Supporting larger transforms, i.e., 64x64 DCT.

Adaptive Transform:

Performed in addition to DCT-II and 4x4 DST-VII, which are employed

in HEVC;

The newly introduced transform matrices are: DST-VII, DCT-VIII,

DST-I and DCT-V.[Bross2016]

Future Video Coding Development (Cont.)


Additional improvements as proposed, for example, by VCEG-AZ07:

Intra-Picture Prediction:

Extended to support 65 intra prediction direction;

0: Planar1: DC

[VCEG-AZ07] And many other tools…

Future Video Coding Development (Cont.)

[Grois2015]


What is Next? – Current Standardization Timeline

[based on

Bross2016]

• Call For Evidence (CfE): Subjective verification of the JEM coding efficiency

compared to HEVC

• Call for Proposals (CfP): Submission and subjective evaluation of new video

coding technologies

20182017

CfP results; and

Starting to work on the

Versatile Video Coding (VVC)

Standard

(Apr., San-Diego, USA)

Final Standard

(October 2020)

Draft CfE

(Jan., Geneva, CH)

2021

Final CfE

(Apr., Hobart, AU)

CfE Results

Draft CfP

(Jul., Torino, IT)

Final CfP

(Oct., Macao, MO)



So, what is next?

The next is the VVC development,

which already started in April, 2018!


Document archives are publicly accessible

http://phenix.it-sudparis.eu/jct

http://ftp3.itu.ch/av-arch/jctvc-site

http://www.itu.int/ITU-T/studygroups/com16/jct-vc/index.html

Overview page:

http://hevc.hhi.fraunhofer.de

HEVC Reference Model (HM) software:

https://hevc.hhi.fraunhofer.de/svn/svn_HEVCSoftware

JVET Joint Exploration Model (JEM) software:

https://jvet.hhi.fraunhofer.de/svn/svn_HMJEMSoftware

JVET VVC Test Model (VTM) software:

https://jvet.hhi.fraunhofer.de/svn/svn_VVCSoftware_VTM

Further Information


Chi2012 Chi, C.-C.; Alvarez-Mesa, M.; Juurlink, B.; Clare, G.; Henry, F.; Pateux, S.; Schierl, T., "Parallel Scalability and Efficiency of HEVC Parallelization

Approaches," Circuits and Systems for Video Technology, IEEE Transactions on, vol. 22, no.12, pp.1827-1838, Dec. 2012.

Cisco2017 “Cisco Visual Networking Index: Forecast and Methodology, 2016–2021”, Online: http://www.cisco.com/c/en/us/solutions/collateral/service-

provider/visual-networking-index-vni/complete-white-paper-c11-481360.html, Cisco Systems Inc., Jun. 2017.

Flynn2015 Flynn, D.; Marpe, D.; Naccari, M.; Nguyen, T.; Rosewarne, C.; Sharman, K.; Sole, J.; Xu, J., "Overview of the Range Extensions for the HEVC

Standard: Tools, Profiles and Performance," in Circuits and Systems for Video Technology, IEEE Transactions on , vol.PP, no.99, pp.1-15, 2015.

Grois2013 Grois, D.; Marpe, D.; Nguyen, T.; Hadar, O., “Comparative Assessment of H.265/MPEG-HEVC, VP9, and H.264/MPEG-AVC Encoders for Low-

Delay Video Applications,” Proc. SPIE 9217, Applications of Digital Image Processing XXXVII, 92170Q, Sept. 23, 2014.

Grois2014 Grois, D.; Marpe, D.; Mulayoff, A.; Itzhaky, B.; Hadar, O., "Performance Comparison of H.265/MPEG-HEVC, VP9, and H.264/MPEG-AVC

Encoders," Picture Coding Symposium (PCS), 2013, pp.394-397, 8-11 Dec. 2013.

Řeřábek2014 Řeřábek, M.; Ebrahimi, T., “Comparison of compression efficiency between HEVC/H.265 and VP9 based on subjective assessments”, Proc. SPIE,

Vol. 9217, Sept. 2014.

Grois2016 Grois, D.; Nguyen, T.; Marpe, D., "Coding Efficiency Comparison of AV1/VP9, H.265/MPEG-HEVC, and H.264/MPEG-AVC Encoders," Picture

Coding Symposium (PCS), 2016, 4-7 Dec. 2016.

Grois2017 Grois, D.; Nguyen, T.; Marpe, D., "Performance comparison of AV1, JEM, VP9, and HEVC encoders," Proc. SPIE, Applications of Digital Image

Processing XL, 6-10 Aug. 2017.

François2016 François, E.; Fogg, C.; He, Y.; Li, X.; Luthra, A.; and Segall, A. "High Dynamic Range and Wide Color Gamut Video Coding in HEVC: Status and

Potential Future Enhancements," IEEE Transactions on Circuits and Systems for Video Technology, vol. 26, no. 1, pp. 63-75, Jan. 2016.

Miller2016 Miller, S.; Nezamabadiand, M.; Daly, S. “Perceptual signal coding for more efficient usage of bit codes”, Annual Technical Conference Exhibition,

SMPTE 2012, pp. 1–9, Oct. 2012.

François2016_2 François, E., Bordes, P., Le Léannec, F., Lasserre, S., and Andrivon, P. Chapter 11 – “High Dynamic Range and Wide Color

Gamut Video Standardization — Status and Perspectives”, in High Dynamic Range Video, edited by Frédéric Dufaux,

Patrick Le Callet, Rafał K. Mantiuk and Marta Mrak, Academic Press, 2016, Pages 293-315.

Nguyen2015 Nguyen, T.; Marpe, D., "Objective Performance Evaluation of the HEVC Main Still Picture Profile," Circuits and Systems for Video Technology,

IEEE Transactions on, vol.25, no.5, pp.790-797, May 2015.

Ohm2012 Ohm, J.; Sullivan, G.J.; Schwarz, H.; Tan, T.K.; Wiegand, T., "Comparison of the Coding Efficiency of Video Coding Standards—Including High

Efficiency Video Coding (HEVC)," Circuits and Systems for Video Technology, IEEE Transactions on , vol.22, no.12, pp.1669,1684, Dec. 2012.

Schwarz2014 Schwarz, H.; Schierl, T.; Marpe, D., “Block Structures and Parallelism Features in HEVC,” in High Efficiency Video Coding (HEVC), Integrated

Circuits and Systems, Eds. Sze, Vivienne; Budagavi, M.; Sullivan, G. J., Springer International Publishing, pp. 49-90, 2014.

SG16−C0806 “Coding Tools Investigation for Next Generation Video Coding,” Qualcomm Inc., Study Group 16, Doc. SG16 – C0806, Jan. 2015.

References


Tsai2013 Tsai, C.-Y.; Chen, C.-Y.; Yamakage, T.; Chong, I.S.; Huang, Y.-W.; Fu, C.-M.; Itoh, T.; Watanabe, T.; Chujoh, T.; Karczewicz, M.; Lei, S.-M.,

"Adaptive Loop Filtering for Video Coding," in Selected Topics in Signal Processing, IEEE Journal of , vol.7, no.6, pp.934-945, Dec. 2013.

VCEG-AZ07 Chen, J.; Chien, W.-J.; Karczewicz, M.; Li, X.; Liu, H.; Said, A.; Zhang, L.; Zhao, X., “Further Improvements to HMKTA-1.0,” Qualcomm Inc.,

Study Group 16, Doc. VCEG-AZ07, Jun. 2015.

Wang2014 Wang, Y.-K., “HEVC and Layered HEVC for UHD Deployments,” Workshop on Media Synchronization for Hybrid Delivery, Oct. 22, 2014,

Strasbourg, France.

Zhang2014 Zhang, X.; Gisquet, C.; Francois, E.; Zou, F.; Au, O.C., "Chroma Intra Prediction Based on Inter-Channel Correlation for HEVC," in Image

Processing, IEEE Transactions on, vol.23, no.1, pp.274-286, Jan. 2014.

DTC2016 “HEVC Products Forecast Overview,” Online: http://www.dtcreports.com/weeklyriff/2016/03/20/hevc-products-forecast-overview/

Bross2016 Bross, B., “Will Standardization do it Again?”, Next Generation Video Codecs, DVB World Asia 2016, Nov. 29 – Dec. 1, 2016, Bangkok, Thailand.

Reinhard2016 Reinhard, E.; Valenzise. G.; Dufaux, F. “Tutorial on High Dynamic Range Video”, EUSIPCO 2016, Aug. 29-Sep. 2, 2016, Budapest, Hungary.

Grois2015 Grois, D.; Bross, B., Sühring, K. “HEVC/H.265 Video Coding Standard (v. 2) Including Range, Scalable, and Multiview Extensions”, Tutorial at

ICIP 2015, Québec, Canada.

Nguyen2018 Nguyen, T.; Marpe, D., “Future Video Coding Technologies: A Performance Evaluation of AV1, JEM, VP9, and HM," Picture Coding Symposium

(PCS), 2018, 24-27 Jun. 2018.

References (Cont.)


Video Coding and HEVC

Dr. Dan Grois, [email protected]
